Performance of k-nearest neighbors algorithm in opinion classification
[ 1 ] Instytut Informatyki, Wydział Informatyki, Politechnika Poznańska | [ D ] doktorant
2013
artykuł naukowy
angielski
- opinion mining
- text mining
- semantic orientation
EN As global Internet network grows rapidly, it is commonly used by a vast number of people to exchange information. By information we mean almost anything, from newspaper articles, to video streaming. One of quite new phenomena is an advent of social network websites, discussion boards (forums), price and product comparators and much more, where users can share their opinions in certain areas. Many such pages implement mechanisms of valuation, where one can, apart from writing a comment, choose whether this comment is positive or negative – in the simplest case. The problem appears when we deal with the text only, without any additional information on the character of the statement, e.g. on discussion boards, or in raw comments to some newspaper article. In this situation the only solution is to process the text, preserving the semantics of the expression in such way, that it can be understood by a computer algorithm. After that, we can evaluate, with a certain probability, whether the processed phrase has a positive or negative value and, therefore, classify it to a positive or negative class of an examined data collection.
97 - 110
CC BY-NC-ND (uznanie autorstwa - użycie niekomercyjne - bez utworów zależnych)
publiczny
15